We will probe the circuitry of spatial working memory using multi single-unit recording in the dorsolateral prefrontal cortex (DLPFC) to study how neuronal activity changes over long delays, and use this information to better understand the circuitry underlying working memory and spatial computations. This project will make use of multi single-unit recording to test several hypotheses regarding the spatial working memory neural architecture. Working memory plays a crucial role in many cognitive functions and as a result has become an important topic of psycophysical, physiological, and computational study. Impairments in working memory are major factors of many disorders such as Alzheimer's Disease and schizophrenia. Understanding the neural mechanisms and circuitry involved in working memory is vital for understanding cognitive functions that depend on working memory as well as for understanding the pathophysiology of related disorders. Spatial working memory decays over time. The neuronal correlates of this decay, measured at the single neuron level, will help elucidate how spatial information is encoded and how spatial working memory functions at the singe neuron and small circuit level. Single cells in DLPFC have circumscribed mnemonic fields, analogous to receptive fields (Goldman-Rakic 1987; Funahashi, Bruce, and Goldman-Rakick, 1989). We will study whether and how these fields change as memory decays. We consider this in the computational context of an attractor network, the premier model for spatial working memory (Compte et al. 2000). Aim 1: Test how the mnemonic fields of individual neurons in DLPFC change as memory decays, and relate these changes at the single cell level to a behavioral assay of memory decay over time. We will obtain tuning curves, in time and space, for the mnemonic fields of individual cells. By recording from multiple cells simultaneously, we will have both absolute and relative information about how these curves change over time, and how these changes relate to trial-by-trial changes in spatial working memory content. From these data, we will determine whether the activity bump formed by the population drifts, drops in amplitude, broadens or thins out over a long delay period. We will also determine which of these changes are correlated with changes in memory decay as reflected in behavioral measures. Aim 2: Test the generality of the spatial working memory mechanisms revealed in Aim 1 by comparing the neuronal correlates of storing a single spatial location with multiple locations. We will construct tuning curves when a macaque is holding two spatial locations in working memory and examine how these tuning curves compare to tuning curves generated when each of the same two locations is maintained individually. We will consider and test hypotheses of a single spatial working memory network vs. multiple independent networks, as well as hypotheses of alternative strategies such as maintaining relative relationships between a set of locations (e.g. remembering a line instead of two locations).